Perly Consulting │ Beck Eco

The State of Play

A living index of AI adoption across industries — where established practice meets the bleeding edge
UPDATED DAILY

The AI landscape doesn't move in one direction — it lurches. Some techniques leap from experiment to table stakes in a single quarter; others stall against regulatory walls, technical ceilings, or organisational inertia that no amount of hype can dislodge. Knowing which is which is the hard part. The State of Play cuts through the noise with a rigorously maintained index of AI techniques across every major business domain — classified by maturity, evidenced by real-world adoption, and updated daily so you always know where you stand relative to the field. Stop guessing. Start knowing.

The Daily Dispatch

A daily newsletter distilling the past two weeks of movement in a domain or two — delivered to your inbox while the index updates in the background.

AI Maturity by Domain

Each dot marks the weighted maturity of practices within a domain — hover for a brief summary, click for more detail

DOMAIN
BLEEDING EDGEESTABLISHED

Presentation creation & enhancement

BLEEDING EDGE

TRAJECTORY

Stalled

AI that creates, structures, and enhances slide presentations from notes, outlines, or spoken input. Includes slide generation and visual enhancement; distinct from event collateral in marketing which produces marketing rather than individual presentations.

OVERVIEW

AI presentation tools have solved speed but not quality or ROI. Gamma ($2.1B valuation, $100M ARR profitably, 70M users as of May 2026) and Microsoft Copilot in PowerPoint (agentic GA April 22 with +25% satisfaction) have achieved production-grade scale and feature maturity. Yet the practice remains firmly bleeding-edge: independent testing shows accuracy failures (76% of Gamma PowerPoint exports require manual fixes, factual hallucination rates 17-44%), NBER's 6,000-executive survey found 80% detecting no productivity impact, and only 3% of Microsoft 365 users converted to paid Copilot, forcing Microsoft to enforce a $30/month paywall in April. The core tension persists between generation speed and production fidelity. Agentic presentation editing now works reliably for multi-step tasks, but brand consistency, export-to-editable-PowerPoint, and audience-aware storytelling remain production blockers. Market capital is routing away from general-purpose presentation tools toward either foundational models or specialized vertical agents, signaling structural headwinds despite user-base scale. Whether tool maturity and user adoption translate into measurable enterprise ROI remains the unsolved question holding this practice at the bleeding-edge tier.

CURRENT LANDSCAPE

The vendor field has consolidated around three tiers. Gamma leads with 70 million registered users and 700,000 daily creations, reinforced by Q2 2026 Series B funding ($68M at $2.1B valuation, $100M ARR profitably) confirming market leadership. Microsoft Copilot in PowerPoint ships embedded across the Microsoft 365 base but requires Designer license for slide generation, blocking students and standard users. Beautiful.ai has carved out an enterprise niche with 100,000+ business customers across 193 countries and documented 3-hour weekly time savings. Prezent claims penetration in 40 of top 50 pharma companies with 70% time reduction metrics.

These adoption signals coexist with widening adoption friction. Microsoft's April 15 decision to remove free Copilot Chat from Office apps and place it behind a $30/user/month paywall reveals a failed adoption bet—only 3% of 450M Microsoft 365 users converted voluntarily to paid tier in six months, forcing Microsoft to restrict free access to drive revenue. Real-world deployment studies show the promise undercut by execution: Workday's survey of 3,200 executives found 37% of time savings lost to rework, with only 14% achieving consistent net productivity gains. Adoption monitoring shows only 35.8% of employees with Copilot access actively use it, compared to 83.1% ChatGPT adoption—a 47-point gap signaling organizational resistance despite enterprise deployment. Accuracy remains a critical barrier: independent testing documents hallucination rates of 17-44% across leading tools, with Gamma inventing fictional competitors, Beautiful.ai misattributing false data to real companies, and Kimi fabricating 78% of financial numbers. Market research projects $4.79 billion by 2029 at 25.4% CAGR, but adoption reality shows 47% of knowledge workers still spend 8+ hours per presentation despite tool availability. The gap between vendor metrics and productivity evidence remains the defining constraint on tier advancement.

TIER HISTORY

ResearchJun-2023 → Jun-2023
Bleeding EdgeJun-2023 → present

EVIDENCE (88)

— May 2026 market metrics: Gamma $2.1B valuation, $100M+ ARR profitably, 70M users, 600k+ paying subscribers, ~52-person team—demonstrates production-grade profitability and category scale.

— Structural repricing analysis: general-purpose AI apps fell from 38% of $100M+ rounds (Q3 2025) to 0% (May 2026). Presentation tools squeezed between foundation models and vertical agents—signals category headwinds.

— Hands-on testing of 12 tools: GenPPT uses Claude/Gemini/ChatGPT with research, generates 10-slide deck in <60 seconds with verified facts. Shows ecosystem diversity and alternative architectures gaining traction.

— Presentation coaching firm critical case study: AI-generated investor pitch deck failed with investors; required human expertise to salvage. Documents limits where AI cannot replace human strategic thinking and judgment.

— Microsoft Copilot Agent Mode GA (April 22, 2026) official metrics: Word +52%, Excel +67% weekly usage, +25% satisfaction in PowerPoint—signals agentic autonomy now reliable for multi-step presentation tasks.

— Practitioner critical test: 19 of 25 Gamma PowerPoint exports (76%) required manual fixes; 6 had visual breaks; conclusion 'best for web-sharing, worst for PowerPoint export'—documents critical production barrier.

— Major product GA milestone: Copilot agents now reliably handle multi-step presentation actions (update slides, apply animations, format autonomously). Reflects year of reliability improvements post-preview.

— Independent hands-on test of new agentic Copilot: realistic gaps in quality, specific failures, user experience friction. Qualifies positive product announcements with empirical limitations assessment.

HISTORY

  • 2023-H1: AI presentation tools emerged with Gamma and Beautiful.AI offering web-based generation; PowerPoint deployed AI features but encountered stability issues; practitioner adoption focused on rapid brainstorming with significant manual refinement required.
  • 2023-H2: Rapid adoption accelerated: Presentations.AI reached 1M users in 84 days; Gamma grew to 10K+ daily signups with millions of active users; enterprise deployments in marketing and corporate sectors; market matured with multiple competing vendors; customization limitations remained the primary adoption barrier.
  • 2024-Q1: Academic research (CHI) advanced AI slide generation from computational notebooks; broad GenAI adoption reached 28% of US workforce; however, deployment barriers surfaced—PowerPoint's Copilot faced reliability and accuracy issues; Gamma's inability to handle branded templates blocked enterprise use; global trust deficits (93% distrust AI outputs for work) and 95% enterprise AI pilot failure rates revealed systemic adoption friction that offset growth from 2023.
  • 2024-Q2: Vendor ecosystem expanded (Decktopus, Magic Slides, Plus AI, Canva Magic Design competing alongside incumbents); however, independent testing revealed persistent quality issues—content mismatches, repetitive layouts, data inaccuracy; generation speed improved (35-39 seconds) but output quality remained below professional standards; market showed growth in downloads but stagnation in production deployment and strategic enterprise use.
  • 2024-Q3: Gamma reached 20M users by September, demonstrating sustained market leadership and user growth despite quality constraints. Duke University practitioner testing validated Gamma's AI image search and generation accuracy for professional use. However, enterprise reliability challenges persisted: NHS experienced a two-week Copilot outage due to code misconfiguration; Microsoft Community reported widespread frustration with Copilot's feature limitations (design, animations, transitions removed or missing) and inability to produce polished presentations. The picture remained consistent: rapid adoption in volume (20M users, multiple vendor ecosystem) but continued friction in production-grade deployment and design fidelity.
  • 2024-Q4: Named enterprise adoption began: dentsu deployed Microsoft Copilot for presentation generation, achieving 15-30 minute daily time savings. Vendor ecosystem continued to expand (Zapdeck and other new entrants launched). Independent practitioner testing by design teams confirmed persistent trade-offs: tools solved speed (20 seconds to minutes) but remained constrained by generic content, limited customization, and inability to handle professional branding. Microsoft repositioned Word-to-PowerPoint export toward AI-driven workflows by replacing it with Copilot, signaling strategic direction but introducing UX friction during transition. Market showed selective adoption in high-volume use cases with broad quality limitations.
  • 2025-Q1: Gamma confirmed market dominance with 10M user acquisition in first nine months post-launch (J.P. Morgan analysis). Beautiful.ai advanced enterprise positioning with Entra ID SSO integration enabling centralized organizational access. Prezent demonstrated B2B adoption across pharma sector (40 of top 50 companies) with named ROI evidence (Cold Chain Technologies 70% time reduction). However, Microsoft Copilot reliability deteriorated—service disruptions in February from code misconfiguration and widespread user reports of task failures and multi-day hang times in March—revealing tension between feature expansion and operational maturity. Market pattern persisted: speed and adoption metrics strong, but customization constraints and operational reliability remained primary friction for enterprise deployment.
  • 2025-Q2: Critical assessments emerged challenging the presentation AI narrative. MIT Sloan expert analysis argued tools save editing time but cannot replace human strategic design, creative judgment, or audience empathy. BBB's National Advertising Division found Microsoft's productivity claims (67-75% users more productive) unsupported, forcing claim modifications. UC Berkeley analysis found 74% of organizations make little AI progress with only 4% generating consistent value; GenAI ROI "approaching zero" with most experiments stalled. Gamma's limitations on contextual understanding (audience, cultural norms, emotional nuance) were documented, and comparative tool reviews noted market growth projections ($572M to $1.1B by 2031) alongside persistent feature trade-offs. Copilot reliability issues continued with user reports of feature failures and task abandonment. The window marked a shift from adoption enthusiasm toward critical scrutiny of actual ROI and tool limitations.
  • 2025-Q3: Feature development continued despite quality scrutiny. Microsoft Copilot in PowerPoint initiated staged rollout with new topic-generation features (rolling out September 2025), addressing prior design-fidelity gaps but with limited geographic availability and deployment barriers. Gamma maintained market momentum with strong community engagement and early-stage adoption, while independent testing reinforced existing trade-offs: tools delivered speed to first draft (minutes vs hours) but failed on brand consistency, creative control, and nuanced storytelling, confirming that hybrid human-AI workflows remained the realistic deployment pattern. Beautiful.ai expanded competitive positioning but user reviews documented persistent export fidelity issues. Market consolidation around Gamma, Microsoft, and Beautiful.ai continued while broader enterprise adoption remained constrained by the speed-versus-customization tension and documented ROI challenges from Q2.
  • 2025-Q4: Market consolidation deepened with clear tier-1 leaders (Gamma 50M+ registered users, 700K daily creations; Copilot auto-installed across Microsoft 365 base) but reliability challenges intensified. Microsoft Copilot in PowerPoint experienced additional service disruptions (November Copilot video-creation function failures affecting UK NHS production deployment, security update triggering image processing bugs). Adoption barriers remained critical: 47% of knowledge workers still invest 8+ hours per presentation despite widespread tool availability, suggesting tool effectiveness or user adoption friction. Individual deployments showed promise (Beautiful.ai enabling freelance marketer $20k gig with afternoon deck creation), but practitioner analysis documented systematic adoption failures in enterprises—vague use cases, insufficient training, and user abandonment when outputs failed to meet expectations. Independent testing confirmed persistent accuracy issues (Gamma incorrectly rendering recent revenue data) and design constraints (standardized templates, limited customization). By Q4, the market pattern was established: tools had achieved scale (100M+ cumulative users, multiple vendors), speed was solved (70% time reduction documented), but production reliability, design fidelity, and user training remained primary friction points blocking broader enterprise adoption.
  • 2026-Jan: Market consolidation continued with growing enterprise deployments contrasting against persistent quality barriers. Mews (global hospitality tech startup with 1000+ properties in 53 countries) deployed Beautiful.ai for sales deck management, reducing update cycles from weeks to minutes—a named-org ROI signal of production maturity in multinational operations. Market research projected the AI presentation generation sector growing to $4.79B by 2029 (25.4% CAGR from $1.54B in 2024), indicating sustained business confidence in the category. However, limitations persisted: ProductLed analysis revealed Gamma users still "delete Gamma's output and hire designers to redo sections," documenting continued quality and customization friction despite market scale. Microsoft Copilot in PowerPoint experienced rollout inconsistencies (Frontier agent missing on web and failing on desktop in January), indicating deployment maturity gaps. Practitioner analysis identified systematic adoption failures: tools produce generic, volume-heavy decks that amplify organizational problems (vague use cases, insufficient training) rather than solving them, explaining why 47% of workers still invest 8+ hours per presentation despite tool availability. The window reinforced the 2025-Q4 pattern: category adoption at scale, speed solved, but ROI, design fidelity, and user enablement remained foundational adoption barriers.
  • 2026-Feb: Critical barriers to production deployment intensified across multiple dimensions. NBER survey of 6,000 corporate executives found 80% detected no discernible productivity impact from AI, with a UK Copilot trial showing zero productivity gain—directly challenging claims that presentation tools deliver measurable ROI. Independent testing revealed accuracy barriers: six major tools (Gamma, Beautiful.ai, Canva, Tome, Kimi) showed factual accuracy rates of 0-44%, with Gamma at 20%, documenting systemic hallucination problems in content generation. Technical barriers persisted: design practitioners identified "Export Problem" where web-based generated slides fail to translate into editable PowerPoint files, and context understanding remained firmly in the human designer's domain. Operational barriers expanded: Copilot in PowerPoint users reported generation failures (fake links, non-functional output) and complex licensing restrictions (Designer license required for slide-generation features, blocking student and standard users). Intermittent regional outages affected Copilot services in late February, further signaling production reliability challenges. By month-end, the practice exhibited classic bleeding-edge tension: massive user bases (Gamma 70M+ users) and vendor ecosystem maturity, but deployment barriers (accuracy, export fidelity, licensing, reliability) preventing broader enterprise adoption. The window marked a consolidation phase where productivity claims faced empirical scrutiny and tool limitations became the primary factor constraining tier advancement.
  • 2026-Mar: Platform consolidation and advanced feature deployment contrasted against critical accuracy failures. Gamma achieved $100M ARR at 70M+ users with approaching 100M user target; Series B funding confirmed $2.1B valuation with superior unit economics vs. competitors (600k paying subscribers, 40% Fortune 500 penetration). Microsoft Copilot in PowerPoint launched significant capability expansion: agentic "Edit with Copilot" mode for complex multi-step tasks, Brand Kits enabling on-brand generation by default, native AI image editing, and model picker for OpenAI/Claude selection. Beautiful.ai achieved $45M Series B funding from General Catalyst with context-aware conversational interface (chat-to-presentation workflow), 100k+ business customers across 193 countries, 100M+ slides created, quantified productivity gains (3 hours/week median, $20k annual value per user). Independent practitioner testing (12-tool comprehensive evaluation) found ecosystem mature with viable alternatives but no dominant tool; all require manual refinement post-generation, particularly for export quality and brand consistency. However, critical accuracy testing documented severe hallucination failures: Kimi fabricated 78% of financial data; Gamma invented fictional competitors with fabricated valuations; Tome inflated metrics 10x; Beautiful.ai misattributed false data to real company names. Feature expansion and organizational maturity indicators strong, but data accuracy and export fidelity remained critical production blockers for financial, analytical, and brand-sensitive deployments.
  • 2026-Apr: Market momentum metrics conflict with adoption reality and ROI uncertainty. Gamma confirmed market leadership with Series B close (April 9): $2.1B valuation, $100M ARR profitably, 70M users, 50-person team. Microsoft, however, signaled adoption friction by implementing April 15 paywall—removing free Copilot Chat from Word/Excel/PowerPoint/OneNote for enterprises >2000 seats, forcing $30/user/month upgrade or loss of in-app assistance. The policy reversal (free in September 2025, paid in April 2026) indicates failed organic adoption: only ~3% of Microsoft 365 users converted to paid tier voluntarily, insufficient to justify continued subsidy. Real-world deployment friction deepened: Workday study of 3,200 executives found 37% of time savings lost to rework cycles and only 14% achieving consistent net productivity gains, challenging vendor ROI narratives. Adoption monitoring metric emerged: only 35.8% of employees with Copilot access actively use it (vs 83.1% ChatGPT), indicating significant organizational reluctance despite enterprise deployment. Accuracy barriers persisted: independent testing confirmed hallucination rates (Gamma 20%, Beautiful.ai 17%, Canva 17%, Kimi 44%, Tome 0%) remain critical for financial/analytical content. Tool performance variance documented: 10x speed range (22 to 228 seconds for 10-slide deck) correlates with quality, suggesting architecture directly impacts user friction. Late-month breakthrough marked significant capability advancement: Microsoft Copilot reached general availability for agentic PowerPoint features (April 22-24 GA), enabling multi-step autonomous actions (update slides, apply animations, format dynamically). Early adoption signals showed +11% tries/user/week and +25% satisfaction in PowerPoint post-agentic launch. However, this feature maturity coexisted with mounting quality concerns: 31% of 500 tested AI-generated decks contained fabricated statistics; comprehensive study of six tools found 20-50% of claims unverifiable, and Gamma users documented continued necessity for manual redesign post-generation. Category maturity evident (100M+ users cumulative, $4.79B projected by 2029), but tier advancement blocked by: ROI gap (productivity claims vs field data), accuracy barriers (hallucination/fabrication rates unsuitable for compliance/financial work), adoption friction (paywall enforced due to low conversion), deployment complexity (37% rework rate negates speed advantage), and critical limits on automation depth (agentic mode advances capability but quality barriers restrict production readiness).
  • 2026-May: Structural headwinds from capital markets combined with confirmed quality barriers at scale. Venture analysis found general-purpose AI apps fell from 38% of $100M+ rounds (Q3 2025) to 0% (May 2026), with presentation tools explicitly identified as squeezed between foundation models and vertical agents—a category-level funding signal. Gamma confirmed continued scale (70M users, 600k+ paying subscribers, $100M+ ARR, 52-person team) but 76% of PowerPoint exports still required manual fixes in independent practitioner testing. Ecosystem diversity grew: hands-on evaluation of 12 tools found GenPPT and other foundation-model-backed entrants generate verified-fact 10-slide decks in under 60 seconds, but all require post-generation refinement. Critical use-case failure persisted: a presentation coaching firm documented an AI-generated investor pitch deck failing with real investors, requiring human expertise to salvage—evidence that strategic narrative and audience judgment remain beyond current tool capability.